ZMIC Journal Club

KAN


张杨
SDS, Fudan University
2024-05-18

KAN
Page 1 of 28

This Saturday, I will present an article about KAN. The author, inspired by the Kolmogorov-Arnold representation theorem, generalized the representation form and constructed a novel type of neural network. Compared to MLP, this network demonstrates stronger accuracy and interpretability. It holds great promise in scientific applications such as solving PDEs and related fields.